Comparing spectros

[Jeff-Grant]

I now have a new 1iPro2 on my io2 and a DTP 70. Everything that I have heard about the DTP 70 is that it is a superior spectro to the humble i1p2. Today I did a quick test on a 21x4 linearity chart. The attached chart from CTP tells the story. The first column is from the i1p2 in spot with two reads, the second is from the DTP, and the third is from the i1p2 in scan mode. The DTP shows a lower number on the deep blacks than the i1p2. That may well be true but I have no idea how to know which one is the more accurate. I would like to think that this is showing me the superiority of the DTP but it would be wishful thinking.

I have discovered the Chromachecker target which I can get for a mere $500. I'm wondering if that is the only choice as I don't want a service, I just want a chart with known values that I can use to compare.

[Mark D Segal]

I now have a new 1iPro2 on my io2 and a DTP 70. Everything that I have heard about the DTP 70 is that it is a superior spectro to the humble i1p2. Today I did a quick test on a 21x4 linearity chart. The attached chart from CTP tells the story. The first column is from the i1p2 in spot with two reads, the second is from the DTP, and the third is from the i1p2 in scan mode. The DTP shows a lower number on the deep blacks than the i1p2. That may well be true but I have no idea how to know which one is the more accurate. I would like to think that this is showing me the superiority of the DTP but it would be wishful thinking.

I have discovered the Chromachecker target which I can get for a mere $500. I'm wondering if that is the only choice as I don't want a service, I just want a chart with known values that I can use to compare.

What you did is showing you the difference between a pair of those columns, not accuracy. You don't say which two were used to derive the last column that contains the dE values. And by the way, for this kind of testing I would use dE(76) rather than dE 2000 - you are looking for simple differences, not differences adjusted for human visual perception.

To do accuracy testing you don't need to buy anything. You need a target with known reference values and you can construct this yourself in Photoshop (see my recent article on Extended Grayscale, this website), and the reference values will be the ones with which you create the patches. The objective of your exercise will then be to discover the differences between the reference values and the printed values read with the spectros. Make sure the reference values you choose are within the gamut of the profile you will be using to print the target. You would use the same printer, same paper, same profile, and make sure the Rendering Intent is set to Absolute. Then read the target with the DTP70. Read it again with the i1Pro2, subtract each of these from the reference values and you will see which set produces the lowest dE(76), and that will be the more accurate one.

If you use Enhanced Matte paper, as you know it is rich in OBAs, hence important with the i1Pro2 to retain the three readings M0, M1, M2 and select the one that produces the most accurate results. It would be preferable to do the testing with a PK paper (wider gamut) that contains no OBAs (eliminates this complicating variable).

Perhaps you aren't aware, because X-Rite is such a dreadful outfit when it comes to documentation, that when the i1Pro2 is is in scan mode, it reads 200 samples per second and averages them. So if you scan the patches at the rate of one per second, you have an average of 200 samples per patch. If you do spot readings, it will sample the whole area of the black hole that the light shines through, taking 200 samples and averaging them.

Given this kind of performance and the fact that there is a 5 years technological gap between these instruments, I would tend to expect the i1Pro2 to stand up very well to a DTP70, both being higher-end professional instruments, but not having worked with the latter I don't know - so the research you are doing is interesting (to see whether new technology means better in this case) and I for one shall be interested in any further results you develop.

[Ethan Hansen]

Mark: Be careful about X-Rite's specs. Yes, the i1Pro2 makes 200 measurements per second. The question is how many of those measurements are valid. The measurement aperture is 4.5mm. Therefore when measuring a target with the default patch size of 8mm, less than half the measurements (44%) are reading the pure patch color. A reason for the ferocious sampling rate is to allow scanning at reasonable speeds while still detecting which measurements should be discarded due to corruption by neighboring patches.

We perform our own in-house calibrations of our aging fleet of Spectroscans. We use BCRA Series II tiles (the standard 12 pack of CERAM tiles plus black and white) as primary reflective standards. We now use the iSis for most measurements and do comparison tests on paper targets between Spectroscan and iSis. Our measurements are in rough agreement with the spec values - calibrated Spectroscans are more tightly grouped than the i1-based iSis (0.3 dE typical/0.8 max vs. 0.4 typical/1.0 max). We saw similar agreement with the spec values for iCColor units when we still used them (aside: if anyone has use for an iCColor or two, let me know). The DTP 70 was comparable to the iCColor in terms of inter-instrument agreement.

Jeff: There can be two causes of black measurements showing darker values. Obviously the blacks really could be darker. Unfortunately this is also a typical sign of the measurement lamp fading. Are you seeing lower L values with both UV included and UV excluded measurements?

Check a 2-D LAB color space plot of profiles made on from the same paper on the i1Pro2 and DTP 70. At L=50, an aging measurement lamp usually produces lower saturation in yellow through orange (gamut boundary at maximum b, a in +/- 50 or so range. A second test is to make prints using the profiles. Soft proof a step wedge in Absolute Colorimetric. Set the steps so you can visually distinguish the difference between adjacent values but not by much. Print each version and check under good illumination if the DTP 70 steps are all visible.

[Mark D Segal]

Ethan - even if only 10 or 20 of those samples are usable, it's still enough to make a useful average for the purposes of eliminating undue impact from outliers. I was just trying to make the point that this isn't a slouchy instrument and is capable of very considerable accuracy.

[Ethan Hansen]

Mark - My contention was not that the i1Pro 2 is not capable. It is, but it also needs a large number of samples to reduce noise. X-Rite has made instruments faster and easier to use but in terms of absolute accuracy in reflectance measurements, none of the current generation are a match for two-decade old instruments.

[Jeff-Grant]

Mark and Ethan, many thanks to you both. How I ended up with the DTP70 is a long and sad tale, due in part to local Xrite folks not being interested in helping me with my Pro2 problem. I can now happily report that, once having found the right people in Xrite, I received stellar support. In desperation, I bought the DTP70 from a friend, having heard nothing but praise for it over the years. I also ended up buying a new Pro2 to use on my io so I now have 2 x Pro2, 1 x Pro, 1 x io and a DTP70. Overkill perhaps for someone doing this mainly for his own pleasure. If anyone is interested, I would be happy to part with a perfectly good Pro2 that won't work on an io at a good price.

I've spent a lot of time digging around with my new/old gear. Doing this, I arrived at the conclusion that I was better off using the Pro2 in spot mode as it would be less challenged with readings. I also increased the default patch size by 1mm, and use 2 reads. This is slow as molasses but is not an issue. I think that I am now getting the best that I can from the io/Pro2 combo.

My concern all along has been that the DTP70 may be showing signs of age. That's why I started out looking for a way of validating what I'm seeing. I'll now start by reading a patch set with UV in and out and report back.

[Jeff-Grant]

I've now come to a screaming halt. I pressed down on the little plastic lever that switches between UV in or excluded and it broke off. It's stuck in limbo so I can't measure anything. I'll report back if I get it going. Thanks again. This definitely isn't the outcome that I was hoping for.

[Ethan Hansen]

Sorry to hear that! I assume the head was positioned in the center of the measuring path? If not, positioning needs to be done through software rather than just shoving the head manually. Might be worth a try to see if the filter lever can be pried with a sharp implement and moved to a definite, final position.

[Jeff-Grant]

Done, I now have an M0 only device, and a query in to Xrite for if they do service and how much.

I dug out a couple of profiles from early on in my trials, one made with the DTP and the other with the i1P. The inner line in the red/orange area is the DTP which looks like what was predicted. I think that I'll wait to hear from Xrite before I make another move.

[Mark D Segal]

Mark - My contention was not that the i1Pro 2 is not capable. It is, but it also needs a large number of samples to reduce noise. X-Rite has made instruments faster and easier to use but in terms of absolute accuracy in reflectance measurements, none of the current generation are a match for two-decade old instruments.

That's interesting Ethan, as you obviously have a professional perspective on it. I've put out for some additional views of the matter and if anything useful turns up I'll mention it here.

[Mark D Segal]

Done, I now have an M0 only device, and a query in to Xrite for if they do service and how much.

I dug out a couple of profiles from early on in my trials, one made with the DTP and the other with the i1P. The inner line in the red/orange area is the DTP which looks like what was predicted. I think that I'll wait to hear from Xrite before I make another move.

For the diagramming it would have been easier to distinguish between the instruments had you chosen to represent each instrument with a different solid colour gamut map. You can do that in CTP. Anyhow, no big deal, one can see regardless that the gamut shapes and volumes are close enough that either instrument would probably do about as well (in respect of print quality) at least in respect of that variable. Profile accuracy however is another matter that cannot be inferred from the gamut shape alone.

[Jeff-Grant]

I'm really interested to hear what others have to say anout the i1p2. For years, i read that the DTP was a stellar device, and that the i1p as it was then was nowhere near as good. When the chancec to get a DTP came up I jumped at it. I am now using multi read spot mode on the io which looks like it should give the best result. My observation of the i1p2 is that it is very good on repeatability.

[Jeff-Grant]

For the diagramming it would have been easier to distinguish between the instruments had you chosen to represent each instrument with a different solid colour gamut map. You can do that in CTP. Anyhow, no big deal, one can see regardless that the gamut shapes and volumes are close enough that either instrument would probably do about as well (in respect of print quality) at least in respect of that variable. Profile accuracy however is another matter that cannot be inferred from the gamut shape alone.

Agreed, Mark, and apologies for not making it more obvious. From day one with the DTP, I wondered why it didn't seem to like darks, and now Ethan has pointed out the problem. On one target, if the first patch was close to black, it would give a weird reading, but if I moved the patch, it was better. I have an awful feeling that the repair price is going to be a deal breaker.

[Mark D Segal]

I'm really interested to hear what others have to say anout the i1p2. For years, i read that the DTP was a stellar device, and that the i1p as it was then was nowhere near as good. When the chancec to get a DTP came up I jumped at it. I am now using multi read spot mode on the io which looks like it should give the best result. My observation of the i1p2 is that it is very good on repeatability.

Much as I have no love-lost for XRite as a company, I have to say - I've been using my i1Pro2 for over a year now, done a considerable amount of work with it as you can peruse from the extensive printer and paper reviews I've researched and written for this website, and I think it is a very good instrument. It bundles the three luminance conditions (M0, M1, M2) in one set of readings and its accuracy/consistency appear to be of a high standard - at the very least high enough to make high quality profiles that deliver high quality prints, and consistent enough in its spot measurements to be analytically very useful. i1Profiler is of course part of the story and a less happy one - it does the needful from a technical perspective, but just a dog's breakfast with no manual.

I have been advised by several knowledgeable people when making profiles to print profiling targets with scrambled patches, so that the differences between them are more obvious to the spectro and therefore less error prone. Using MeasureChart in Spot mode does work very well for measuring various patch sets (making sure the patches are large enough for accurate unambiguous placement of instrument) or individual colours, but this of course is not for creating profiles.

As for servicing whic you mention just above, There are two hurdles with XRite. In their "infinite wisdom" they do choose rather aggressively when to discontinue support to "legacy" products, so the first hurdle is whether they will service at all, and then of course the cost.

[Jeff-Grant]

I'm in violent agreement with you on all of that. I'm still contemplating whether to use DropRGB but, despite the lack of documentation i1p does build good profiles. Using the io let's me do the spot read easily. I woukd not contemplate doing that hy hand. I have been using scrambled patches for years, not as pretty but seems to work well. I also up the patch size by at least 1mm from the default.

[Rhossydd]

To do accuracy testing you don't need to buy anything. You need a target with known reference values and you can construct this yourself in Photoshop

I'd agree that to do any sort of absolute accuracy testing with spectros needs a known reference target, but building a target yourself seems too unreliable.
It assumes that the profile is absolutely correct and that the printer can deliver output absolutely consistently. They are both big assumptions.
In my experience printers don't deliver results more accurately than spectros.

Comparing different spectros for absolute accuracy is a complex and difficult task, especially so if you need specially formatted targets for autospectros.

[Mark D Segal]

I'd agree that to do any sort of absolute accuracy testing with spectros needs a known reference target, but building a target yourself seems too unreliable.
It assumes that the profile is absolutely correct and that the printer can deliver output absolutely consistently. They are both big assumptions.
In my experience printers don't deliver results more accurately than spectros.

Comparing different spectros for absolute accuracy is a complex and difficult task, especially so if you need specially formatted targets for autospectros.

It depends on what you are testing and to what end.

[Rhossydd]

It depends on what you are testing and to what end.

The OP's original question is fairly simple;
"I have no idea how to know which one is the more accurate"

To know the answer to that requires comparing to a known accurate reference.

DIY charts have a use for comparative testing, but for absolutes they won't be any help.

[Mark D Segal]

Chasing absolute accuracy with any of this stuff is somewhat ephemeral, because everything in the chain can have small variances. Comparative testing suffices for the OP's purposes and that's the context I had in mind when I posted my response. I.E., amongst the options he has on hand, which is going to deliver the most correct and reliable outcomes would seem to be the practical question at hand.

[Rhossydd]

Comparative testing suffices for the OP's purposes

Well that might be your opinion, but his question isn't answered; "which one is the more accurate"

The only way to discover that is working from a known reference.
A CC24 might help get an insight, a Colorchecker SG might get a bit further, but there will still be a need to do some comparative testing to discover the repeatability of instruments to know what the limitations of the spectros measurements are.

I understand the interest in this issue having been through these enquiries myself when I first owned more than one spectro. After lots of testing and analysis my own opinion is that when the numbers start to drop below 1dE it's time to stop worrying about it and get on with making pictures. The CM tech will be good enough.